José Antonio Rabadán Rubio

Modeling a risk detection system for elderly's home-care with a network of timed automata

We address the problem of modeling a risk detection system oriented to elderlys home-care, which may cause the generation of an alarm if there is evidence of a serious problem. We have chosen a network of timed automata as a suitable formalism to solve the problem in a simple and elegant fashion. Our model is capable of representing the subjects environment, modeling his behavior dynamics, detecting several types of abnormal behavior patterns, and deciding whether they have enough evidence of being dangerous to generate an alarm. In addition, the formalism allows concurrent activity monitoring in different rooms, what is useful to determine if the person is alone at home, in order to avoid the generation of false alarms when the elder is supposed to be already under human care. We have verified critical properties of the model by using a model-checking engine.